Re: [PATCH 3/3] Implement the notStopped DAP response

["Kévin Le Gouguec" <legouguec@adacore.com>]

Tom Tromey <tromey@adacore.com> writes:

> @@ -21,8 +21,17 @@ from .startup import exec_and_log, in_gdb_thread, log
>  from .modules import is_module, make_module
>  
>  
> +# True when the inferior is thought to be running, False otherwise.
> +# This may be accessed from any thread, which can be racy.  However,
> +# this unimportant because this global is only used for the
> +# 'notStopped' response, which itself is inherently racy.
> +inferior_running = False
> +
> +

Out of curiosity, could `inferior_thread ()->state` have helped here?
(Exposing it as a gdb.Inferior method perhaps?)

Not necessarily with the "racy" part; thinking more of the "keep a
single source of truth" part (i.e. dispense with relying on event
notifications).

> +# A helper decorator that checks whether the inferior is running.
> +def _check_not_running(func):
> +    @functools.wraps(func)
> +    def check(*args, **kwargs):
> +        # Import this as late as possible.  FIXME.
> +        from .events import inferior_running
> +
> +        if inferior_running:
> +            raise Exception("notStopped")
> +        return func(*args, **kwargs)
> +
> +    return check

Assuming this is what the FIXME is about (apologies if not): an
alternative could be doing a bare `import .events`, then lookup
`.events.inferior_running` in check().  Or defining an accessor function
in .events and importing that instead of this global.

> @@ -178,6 +199,11 @@ def request(name: str, *, response: bool = True, server: bool = False):
>  
>      If SERVER is True, the function will be invoked in the DAP thread.
>      When SERVER is True, RESPONSE may not be False.
> +
> +    If MUST_BE_STOPPED is True (the default), then the request will
> +    fail with the 'notStopped' reason if it is processed while the
> +    inferior is running.  When MUST_BE_STOPPED is False, the request
> +    will proceed regardless of the inferior's state.
>      """
>  
>      # Validate the parameters.

(Feeling like EXPECT_STOPPED might better convey that parameter's role
as a precondition, rather than as a "call to action" for request(); tiny
nit though, the doc clarifies the intent anyway)